Dreams Come True. IMDB: Are you worried or excited?
I have been busy working on several projects for the last couple of months. At last I am able to share some insights with the community.
Among other things I have made an inroad into the In-Memory DB technology. The technology is far from being new. Unfortunately, it is still under-utilized and under-appreciated. One of the reasons is that it is often treated as an alternative for traditional OLTP DBMS. While it definitely may be used as an OLTP altrenative, the worries it brings with it will out-weight its performance benefits and thus result in the solution being dismissed as “second best.”
As with other technologies (e.g. Replication vs. Clustering I have briefly visited last times), if it is applied to a wrong use case its benefit will definitely be minimal, if at all. It is very important to understand technology and apply it to situations that suit it most. Otherwise it will be misused, misappreciated and ultimately forgotten…
IMDB is not an exception to the rule. While it represents a technological leap into boosted performance (In-Memory will inevitably be faster than Through-the-Cache solution), the worries it bring with it if applied thoughtlessly will only do it harm.
I will not talk in detail about what Sybase IMDB is here. Briefly, it is an in-memory solution provided by Sybase which is available out-of-the-box, operates like just another database within your ASE server (~tempdb?), understands the same syntax and demands the same knowledge you probably already have as Sybase DBA/programmer. It may boost your performance officially tens and unofficially hundreds of times (empirical data), while demanding from you only the suitable hardware (OS memory) and a license.
When I have played with it in my developer edition, I have been thrilled (loading gigs of DB in seconds? building clustered indices on 1m+ rows in seconds? significantly cutting maintenance / performance testing time, &c). When presented the solution to customers, and running controlled POC-s, the customers started to apply it to imaginary use-cases even before I have finished presenting the solution. If you have vital imagination and do not think of IMDB as a replacement to the traditional OLTP database, you will find this technology very attractive. If you manage your IT resources with care, you will find it not expensive either. Whether you will be worried by the loss of the ACIDity of [one of ] your DBMS or you will be excited about the “extreme” performance possibilities the technology open for you depends on the choices you make.
You may read my presentation on it here: IMDB.
As to me, any decent IT with DWH/DSS orientation must have this technology available. It depends if you are oriented toward your future or are still worried about your past. Reminds me of some famous quotation by Nietzsche…
I would recommend testing this technology first hands. Testing it costs nothing anyway…